A distilled multi-order kinematic model outperforms pose-based baselines on a synthetic classroom incident dataset while using one-tenth as many parameters as PoseC3D¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? On the authors' synthetic classroom incident dataset, the proposed method reaches 71.78% accuracy, improving over PoseC3D by 1.24 percentage points and over MSG3D by 2.90 percentage points. → Paritosh Parmar 2026 q2 i? The proposed model uses only one-tenth as many parameters as PoseC3D while achieving superior performance on the synthetic dataset. → Paritosh Parmar 2026 (2)

Evidence¶

Paritosh Parmar 2026¶

Paritosh Parmar, Landy Lan, Hong Yang, Chen Yi, Chiat Pin Tay. (2026). Robust and Efficient Motion Reasoning for Privacy-Aware Classroom Incident Recognition. arXiv preprint. https://arxiv.org/abs/2608.05115 doi:10.48550/arxiv.2608.05115

q2 · i? · design · r2

Comparison experiment on the authors' synthetic classroom incident dataset (80/20 train-test split) against AAGCN, STGCN++, CTRGCN, MSG3D and PoseC3D. The proposed method "achieves the highest accuracy of 71.78%"; accuracies are reported without effect sizes.

"Our method achieves the highest accuracy of 71.78%, outperforming all compared baselines. In particular, it improves over PoseC3D, the strongest baseline on this dataset, by 1.24 percentage points, and over MSG3D by 2.90 percentage points."

Paritosh Parmar 2026 (2)¶

Paritosh Parmar, Landy Lan, Hong Yang, Chen Yi, Chiat Pin Tay. (2026). Robust and Efficient Motion Reasoning for Privacy-Aware Classroom Incident Recognition. arXiv preprint. https://arxiv.org/abs/2608.05115

q2 · i? · design · r2

Efficiency comparison reported alongside the synthetic-dataset results and plotted in Figure 2 (performance vs GFLOPs, circle diameter indicating model size in million parameters). The model "uses only one-tenth as many parameters as PoseC3D"; no effect size is printed.

"Notably, our model is computationally efficient: it uses only one-tenth as many parameters as PoseC3D, yet achieves superior performance (Figure 2)."

Discussion¶

Related Claims¶

• Classroom incident recognition is substantially harder than established general action-recognition benchmarks, with PoseC3D dropping from over 94% accuracy to 70.54% — related
• Distilling a multi-order hierarchical teacher into a smaller zeroth-order student lets the student outperform the teacher, and improves 4.1 percentage points in accuracy over STGCN++ — related
• Models trained on synthetic classroom incident data transfer zero-shot to real-world footage, with the proposed method best at 63.41% though a synthetic-to-real domain gap remains — related
• The framework is robust to 15% ASR transcription noise and retains 99.6% of its ICS under zero-shot cross-domain transfer to EdNet-Math — related